Duolingo sheds some human workers as AI threatens to upend the $65 billion translation industry

Lost in translation? Or lost in automation? One thing is for sure, the translation industry is on the precipice of a massive change, in large part due to the advent of more powerful and ubiquitous artificial intelligence tools.

Late last month, a Reddit user claimed that the popular language-learning app Duolingo had “offboarded a huge percentage of its contractors who did translations.” The user added that “this is because they figured out that AI can do these translations in a fraction of the time,” leading to cost-savings for the company. Some remaining team members “will just review AI content to make sure it’s acceptable,” the Redditor added, claiming to have worked for the app for five years.

While Fast Company was unable to directly verify the specific user’s claims, Duolingo did confirm that the company had offboarded contractors at the end of 2023. It was “a small minority” of its overall contractor base, according to a Duolingo spokesperson, and “the vast majority were retained.”

Duolingo also said that the reasons those contractors were let go were that their contracts expired, and that the company had evolved its “content creation operations and no longer need as many people to do this work.” The company also said that it does utilize AI to a certain extent, including in course content creation. 

The average person may not give much thought to the world of translation, but it’s big. Numerous companies specialize in translation services, such as TransPerfect, LionBridge, and Keyword Studios, and it’s a roughly $64.7 billion industry worldwide. And while Duolingo may not be a translation company per se, its business is largely built on translation—accurate translation. The company’s larger focus on AI tools may be a sign of what’s to come for the larger translation industry.

“Things are changing at a very quick pace”

Alon Lavie, vice president of AI research at Phrase, a translation automation platform, says that combined, the industry employs “several hundred thousand” human translators and a majority of them focus on translating enterprise content from a source language into their native target language.

“The entire industry is built on a foundation in which content is generated in one language, and then translated into many other languages,” Lavie says. But the widespread adoption of generative AI tools over the past year, which has affected many other industries, is threatening that foundation, and the translation industry as a whole, he says.

“The floodgates opened by OpenAI a year ago really exposed not just the general public, but particularly the translation industry at large to the potential recent breakthroughs in AI technology,” he says. “It’s happening just like it’s happening in a lot of other areas and verticals—things are changing at a very quick pace.”

In that sense, Lavie says that the translation industry itself is likely to get more efficient, but less human. “This is an industry that’s been around for a long time, it’s been very human,” he says. “Human translation has always been a bottleneck because it’s slow and expensive and difficult to scale up.”

New AI tools are helping automate the translation process. And, as in other industries, a translator who can leverage those tools and implement them into workflows to become faster, more efficient, and more cost-effective, is maybe more likely than others to make it through the coming round of disruption employed.

Others in the industry say that, interestingly enough, the translation industry is one that’s been at the forefront of the AI revolution for perhaps longer than any, or at least longer than most.

“We’ve been dealing with this for a lot longer than most people think,” says Ben Karl, board director at the American Translators Association (ATA), who also owns a translating service. Karl says that there have been oft-used translation tools on the market for a long time, such as Google Translate, but it’s really been the past several years that things have picked up as it relates to AI use in translation.

For instance, he points out that DeepL, a German neural-machine translation service, has been one of the front-runners in the industry, having launched its platform in 2017. “They really pioneered the use of machine translation,” Karl says. “Since then, a lot of translation companies, mostly large agencies, sell a service called ‘post-edited machine translation’—there are a lot of translators who’ve been providing a service that’s essentially copy editing AI-generated text.”

A game with high stakes

Karl says that there are risks and downsides to AI-powered translation tools, which some companies that traditionally relied on human translators may ignore in lieu of cost savings.

Specifically, languages contain multitudes of quirks that could easily trip up an AI translator, while a human would be able to pick up on them. Dialects or phrases that are specific to certain regions—how the word “dinner” may refer to different meals depending on what part of the country you’re from, for example—may not be intuitive to an AI tool. There could also be generational differences at play.

But the real danger is that incredibly important information that needs to be clearly conveyed could be mistranslated with disastrous results. For instance, if a company that prints labels for medications accidentally uses the word “do” instead of “don’t” on a label, the implications could be far-reaching, and even deadly. Or, as reported by The Guardian last year, reliance on translation services and tools instead of human translators is creating problems for asylum-seekers and refugees, who are being turned away from the United States due to bad or inaccurate translations.

As such, while AI is likely to have a massive effect on the translation industry, humans still have to play a key role in the immediate future.

“We’re in an interesting moment where we’re all trying to figure out how to adapt to this tech,” says Karl. “Translators are uniquely poised because we’ve dealt with this for a while already. It’s been a part of a lot of translators’ workflows and toolboxes for a long time. But this is like it’s on steroids,” he says. 

Lavie says that AI could lead to the translation industry being almost unrecognizable within several years. “The dynamics of the industry are in upheaval,” he says, and many incumbent companies are having “the rug pulled out from underneath their feet. 

“In five years, the industry will look very different,” he says.